Comparaison de résumés linguistiques

When tabular data cannot be directly mined, due to their size or for privacy reasons, their summary may still be available for analysis. The approach proposed in this paper provides users with a linguistic description of the  data changes between the fuzzy linguistic summaries of two datasets. A first strategy processes exhaustive summaries containing one sentence for each of the subspaces that can be formed using terms from the vocabulary. A second strategy is proposed for condensed summaries, that involve informative sentences only. Experimentation conducted on artificial datasets confirm the relevance of this second strategy in terms of computational cost and informativity of data changes that can be tracked.